Configurable Safety Tuning of Language Models with Synthetic Preference Data
Paper
•
2404.00495
•
Published
•
2
Configurable Safety Tuning ⚙️
CST allows for configurable inference-time control of LLM safety levels, so users can dictate model behavior based on the system prompt
-
-
vicgalle/configurable-system-prompt-multitask
Viewer • Updated • 1.95k • 165 • 19 -
vicgalle/ConfigurableHermes-7B
Text Generation • Updated • 4.37k • 3 -
vicgalle/ConfigurableSOLAR-10.7B
Text Generation • Updated • 3.73k • 2 -
vicgalle/ConfigurableBeagle-11B
Text Generation • Updated • 4.36k • 3 -
vicgalle/Configurable-Llama-3-8B-v0.3
Text Generation • Updated • 4.23k • 16 -
vicgalle/Configurable-Hermes-2-Pro-Llama-3-8B
Text Generation • Updated • 9.33k • 6 -
vicgalle/Configurable-Yi-1.5-9B-Chat
Text Generation • Updated • 5.59k • 2 -
vicgalle/Configurable-SmolLM-1.7B
Text Generation • Updated • 12 • 1 -
vicgalle/Configurable-Llama-3.1-8B-Instruct
Text Generation • Updated • 2.82k • 11 -
vicgalle/Configurable-Hermes-3-Llama-3.1-8B
Text Generation • Updated • 2.45k • 4